Research map / Large language models / Retrieval and knowledge
Query-Efficient Domain Knowledge Stealing Against Large Language Models
Zhengao Li, Xiaopeng Yuan, Bolin Shen, Kien Le, Haohan Wang, Xugui Zhou, Shangqian Gao, Yushun Dong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add chain-of-questioning (CoQ) (ablation) | query-efficient stealing framework · MedQA | accuracy: 59.1 -> 62.0 | Improved |
| add chain-of-questioning (CoQ) (ablation) | query-efficient stealing framework · FOMC | accuracy: 52.2 -> 54.0 | Improved |
| add feedback-guided scheduling (FB) (ablation) | query-efficient stealing framework · MedQA | accuracy: 58.9 -> 62.0 | Improved |
| add feedback-guided scheduling (FB) (ablation) | query-efficient stealing framework · FOMC | accuracy: 53.9 -> 54.0 | Improved |
| add cluster-guided exploration (ablation) | query-efficient stealing framework · MedQA | accuracy: 60.8 -> 62.0 | Improved |
| add cluster-guided exploration (ablation) | query-efficient stealing framework · FOMC | accuracy: 53.4 -> 54.0 | Improved |
| use structured exploration instead of EvoKD | LLaMA-3.1 8B surrogate · FOMC | accuracy: 43.5 -> 54.0 | Improved |
| use structured exploration instead of Model Leeching | LLaMA-3.1 8B surrogate · ChemProt | accuracy: 44.7 -> 43.1 | Hurt |
| use structured exploration instead of Model Leeching | LLaMA-3.1 8B surrogate · PubMedQA | accuracy: 74.3 -> 74.9 | Improved |
| use structured exploration instead of Model Leeching | Gemma-2 9B surrogate · FPB | accuracy: 60.1 -> 68.5 | Improved |
| use structured exploration instead of EvoKD | Gemma-2 9B surrogate · HeadLine | accuracy: 78.5 -> 80.4 | Improved |
| train surrogate on 500 stolen QA pairs vs untuned base | LLaMA-3.1 8B · ChemProt | accuracy: 24.7 -> 43.1 | Improved |
Closest papers by meaning
- BLADE: Enhancing Black-Box Large Language Models with Small Domain-Specific Models (AAAI 2025 · similarity 0.88)
- Physics of Language Models: Part 3.1, Knowledge Storage and Extraction (ICML 2024 · similarity 0.87)
- Physics of Language Models: Part 3.2, Knowledge Manipulation (ICLR 2025 · similarity 0.86)
- Iterative Self-Incentivization Empowers Large Language Models as Agentic Searchers (NeurIPS 2025 · similarity 0.86)
- Structural Entropy Guided Agent for Detecting and Repairing Knowledge Deficiencies in LLMs (NeurIPS 2025 · similarity 0.85)
- Once Read is Enough: Domain-specific Pretraining-free Language Models with Cluster-guided Sparse Experts for Long-tail Domain Knowledge (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
